The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external video links, transcripts, or user-provided descriptions to perform structural analysis, which could contain malicious instructions.
Ingestion points: The skill accepts video links or text transcripts in the 爆款来源 (Viral Source) field.
Boundary markers: The skill uses structured Markdown tables and JSON templates to organize its output, providing some separation.
Capability inventory: The skill is limited to text and JSON generation. It does not perform network operations or file writes directly, though it provides commands for a user to run locally.
Sanitization: Includes a 'Compliance' section with a checklist to filter out prohibited advertising terms and copyright-infringing content.
[EXTERNAL_DOWNLOADS]: The skill references external URLs for image generation and code repositories.
Evidence: Mentions gptimage2.asia as a resource for generating missing images and github.com/xianyu110/ecommerce-video-skills as a source repository. These are documented as part of the intended workflow.

viral-remake